Currently, load sharing is generally performed based on 5-tuple information. For example, a hash operation is performed on an IP source address, an IP destination address, a protocol number, a source port, and a destination port in a 5-tuple, and a signal stream is exported to different next-hop routing devices according to a result of the hash operation. Generally, such hash is statically preconfigured on a network device, such as a router, and cannot be adjusted according to a specific condition.
As service types of operators increase, a more flexible service chaining technology is put forward. This technology provides a basis for flexible service processing by linking multiple service processing functions. In the service chaining technology, a service classifier (SCLA) classifies a service stream, and then a service forwarding entity (SFE) forwards the service stream to different service processing entities (SPE) for service processing. In the service chaining technology, service load sharing cannot be dynamically adjusted according to usage of a service network resource. Therefore, the service chaining technology can support only a small-size service network and is not applicable to a service environment in which a large quantity of services and a large quantity of service processing function units exist.